| | |
Summary: ENTROPY-BASED VARIABLE FRAME RATE ANALYSIS OF SPEECH SIGNALS AND ITS
APPLICATION TO ASR
H. You, Q. Zhu and A. Alwan
Electrical Engineering Department, UCLA
Los Angeles CA90095, USA
hyou, qifeng, alwan@icsl.ucla.edu
ABSTRACT
Most speech processing algorithms analyze speech signals
frame by frame with a fixed frame rate. Fixed-rate analy-
sis is inconsistent with human speech perception and effec-
tively assigns the same importance or `weight' to all equi-
duration frames. In [1], we proposed a variable frame rate
(VFR) analysis technique that is based on a Euclidian dis-
tance measure. In this paper, we propose another approach
for VFR based on the entropy of the signal. We compare
entropy and Euclidian distance measures for VFR in ASR
experiments using the Aurora2 and TI46 databases. Better
performance is observed for the entropy-based VFR over
our earlier VFR approach and over the fixed-rate system.
1. INTRODUCTION
|
